For the evaluation of the performance characteristics of a communication path, e.g. the bit error rate, the number of packets lost and/or inserted, the delay jitter, etc. . . . , a "measurement traffic" of test packets has to be generated. This measurement traffic is either constituted by test packets added to the existing live traffic of data communication packets or by test packets simulating such live traffic.
In case of addition of test packets, the ratio of these test packets on the live packets must remain very low, e.g. less than one percent, to avoid too much disturbance on the performance characteristics of the live transmission. As a result, a long measurement time is required for establishing the packet loss since the latter is generally of about 10.sup.-8 for a switching network and is calculated by means of the test packets only.
In case of simulation, real time traffic generators or sources are simulated by mathematical models, e.g. by Gaussian or exponential distributions, and to obtain reliable results the simulated test packet stream must resemble as much as possible the live traffic. This resemblance is however not obvious to obtain in practice.